Dear Dojo Community Members,
We received a question from another DataMiner user and it seems like a good case to ask on Dojo to let the knowledge be shared.
The user is administrator of a DataMiner System, that has 16 active agents (4 of them are in a failover pair, so 20 agents in total). For reading and writing data, there is:
- A Cassandra cluster of 8 nodes (Linux machines, Ubuntu OS 22.04.3 LTS, Cassandra version 4.1.3). It has a Replication factor of 3 and the consistencyLevel is set to "Quorum"
- A OpenSearch cluster of 4 nodes (also Linux machines, Ubuntu OS 22.04.4 LTS, OpenSearch version 2.14). It has a replication factor of 2.
The user wants to do good maintenance on the Linux Servers, and for example do security patches on a regular basis. Therefore, he would also like to run the "apt upgrade".
Is it correct to say that as long as the upgrades are done separately, in a sequence, and only one node is updated and reboot at a time, that DataMiner should not be stopped as all data is replicated?
Are there other important things/steps to keep in mind?
Kind regards,
Joachim
For Cassandra:
- A cluster of 8 nodes and replication factor of 3 will handle shutting down of one node with no impact. It is therefore safe to do a rolling upgrade one node at a time. DataMiner should not experience any impact, although warnings about Cassandra cluster health may be visible.
- The upgrade procedure, however, is more complex than just running "apt upgrade".
See DataMiner documentation: Updating Cassandra | DataMiner Docs
And DataStax documentation: Upgrade an online cluster | Luna for Apache Cassandra | DataStax Docs
For OpenSearch:
- A cluster of 4 nodes and replication factor of 2 should tolerate shutting down of one node with no impact.
- Rolling upgrade of an OpenSearch cluster is possible within a major version range (e.g. 2.x): Rolling Upgrade - OpenSearch Documentation
See also DataMiner documentation: Securing the OpenSearch database | DataMiner Docs
Thank you very much Alexander!